The role of suprasegmental cues in perception of sentences with linguistic ambiguity under informational masking.

Introduction: Real-life communication contains rich and informative suprasegmental cues, such as variations in intensity, duration, and fundamental frequency. Although suprasegmental information is an essential component of spoken communication, we know little about its role in speech perception in the presence of background masker. Building on literature showing that suprasegmental cues facilitate the processing of spoken sentences with linguistic ambiguity, we addressed two questions in the present study. First, does the facilitative effect of suprasegmental cues on speech recognition interact with the amount of informational masking in speech maskers? Second, how do listeners use suprasegmental and lexico-semantic cues when listening to linguistically ambiguous sentences masked by competing speech maskers? Methods: We collected both offline performance data (recognition accuracy) and online processing effort data (pupil dilation) from 37 young adults with age-typical hearing. The speech material consisted of 15 sets of temporarily ambiguous early closure sentences, each with two suprasegmental conditions (facilitative vs. neutral) and two lexico-semantic conditions (transitive vs. intransitive subordinate verb). These sentences were embedded in original and time-reversed two-talker speech maskers differing in the amount of informational masking. Results: Recognition accuracy was higher with facilitative suprasegmental cues, particularly in maskers with less informational masking, as well as with facilitative lexico-semantic cues. Listeners expended greater processing effort throughout the sentence when suprasegmental cues were neutral, especially under more adverse conditions (i.e., stronger informational masking or greater linguistic ambiguity). Discussion: This study makes multiple contributions to the literature. First, the recognition accuracy data showed that informational masking and linguistic ambiguity interact with suprasegmental effects: these adverse conditions reduce the benefit of facilitative suprasegmental cues for speech recognition. Second, under stronger informational masking and greater linguistic ambiguity, the absence of facilitative suprasegmental cues increased effort during online speech processing. Third, we found that facilitative suprasegmental cues improved immediate recall of segmental information (i.e., words) in speech perception in speech maskers. Finally, our accuracy and effort data demonstrate the importance of using both offline and online measures of speech processing, as each reveals different aspects of the dynamic process of speech perception under adverse conditions.